Flaky test detection
Flaky tests are eroding trust in your test suite — people rerun red builds without reading them. Walk me through an agent that hunts flakes: how it detects them, whether it's allowed to quarantine a test on its own or must open a ticket for a human to approve, and what evidence it attaches. How would you prove it's finding real flakes and not masking genuine bugs before the team relies on it? Finish by selling me the idea in a couple of sentences.
find_quarantine_candidates(run_records: list[str]) → list[str][["login_flow,c1,pass","login_flow,c1,fail","checkout_sum,c1,fail"]]out["login_flow"][["auth_token,c1,pass","auth_token,c2,fail"]]out[][["cache_evict,c9,fail","cache_evict,c9,fail","cache_evict,c9,pass"]]out["cache_evict"]Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.